The Research Of Performance Model For Multiple Components Based Web Application

As long with the progress of the computer science and the Internet, Web application are used in almost every industries. Especially in recent years, people want the web application with more and more useful function, which lead to the scale of the web system keeping to enlarge and increasing complexity. That is the reason why make sure the Web application on the good reliability and performance is so important, and performance testing is a indispensable part of Web application testing. Performance testing is conducive to performance problems early discovery, the discovery process can be time performance problems and system bottlenecks, during performance testing, Web service system through performance optimization, and ultimately to achieve greater system availability, reliability and performance Reduce the decline in service quality and therefore bring a variety of losses. Web application performance testing is different from ordinary client/server application on many aspects:the load uncertainty, complex structure, the composition of various different operating mechanisms, inconsistent operation. This paper presents a new approach to quantify the performance of multiple components based web application. We use components analysis driven performance model which include components resource need profile, remote component invocation profile and inter-component communication pattern using a kernel tools without the need of knowing details of the application. Our model differentiates remote invocations from fast-path calls between co-located components and we measure the network delay caused by blocking inter-component communications. Our model can accurately predict the web application throughput and the request response time as long as the component placement is given. In addition to supporting static component placement, our model also could be used to guide dynamic component migration for achieving better performance. Component migration requires up-to-date knowledge of runtime dynamic workload characteristics. This performance model could be used to assist the following system management functions: optimized component placement, capacity planning, and cost-effectiveness analysis. Finally, we have a summary of the work in this paper. The last part in the paper, a outlook of the technology and application in the paper is given, and point out our future research work..